無需紅外光源之即時眼動、眨眼與視線偵測系統

A Real-time Eye Localization, Eye Blinking Detection, and Gaze Detection System without Infrared Illumination

陳柏淳(Bo-Chun Chen)

指導教授 ： 簡韶逸

國立臺灣大學/電機資訊學院/電子工程學研究所/碩士(2014年)

https://doi.org/10.6342/NTU.2014.10047

摘要

在過去的40年中，有非常多的研究投入於建造一個健全且便宜的眼動儀，並將此系統用在人機介面上。然而，要將眼動偵測、眨眼偵測與視線偵測全部整合成一個有效的人機介面系統仍然非常的困難。最主要的難處在於如何降低這類系統的成本與提高其可用性。根據每個人在生理的不同，其眼睛在臉上的幾何關係、位置甚或是光影變化都會有所不同。過去的眼動儀因為這些限制常常必須依賴一些額外的裝置來提升準確度，例如，眼電圖法中使用的電極貼片、搜尋線圈法中使用的大型環繞磁場建造裝置、基於攝影機之眼動追蹤系統中的紅外線光源以及多個立體攝影機。為了要建造一個便宜、好用、且對使用者無負擔無需額外累贅的眼動儀，我們發展了一個不但可以即時偵測使用者視線且可以即時偵測使用的眨眼與眼動的視線偵測演算法。我們的系統只需要兩台網路攝影機，一台捕捉場景影像，一台捕捉使用者的眼睛影像。我們也自己製作了一個可以搭載如上所述的兩台網路攝影機的眼鏡。考量到若是架設紅外線光源並且於如此近的距離照射眼睛恐有健康上的疑慮，於是我們決定將紅外線光源移除，整個系統只接收環境的自然光源而無其他額外光源。即使我們的系統有著相對較差的光源環境且直接地影響到了系統的輸入，我們的效能上並不會輸給有紅外線光源的系統，我們甚至可以與一些經典的系統有著相近的表現。在眨眼偵測上，意識性眨眼可用於一個人機介面的控制訊號，我們在意識性眨眼上針對四個受測者做了實驗並且得到了96.8%的準確率。

關鍵字

眼動儀 ； 眼電圖法 ； 搜尋線圈法 ； 眼動偵測 ； 眨眼偵測 ； 視線偵測

並列摘要

During the last 40 years, tremendous efforts has been made to develop robust and cheap eye tracking systems for human computer interface. However, eye detection, eye blinking detection, gaze estimation, and integrating all of these into a useful interface system remains challenging.The primary difficulty of such work is the availability of a reliable and low-cost eye tracking system. Due to the individuality of eyes, occlusion, variability in scale, location, and light conditions, past eye tracking systems always needed to rely on extra device to help improving the overall accuracy, such as electrodes in Electrooculography(EOG) system, coil and surrounding filed frame by search coil method,infrared light sources, and multiple stereo cameras in camera-based eye trackers. Towards making an eye tracking system which is cheap, easy to use and available to interface users without redundant device, we have developed a gaze estimation algorithm that not only detect user’s gaze, but also the location and blinking of user’s eye in real-time. The only devices we use for input are a scene web-cam, which is used for capturing scene images, and an eye web-cam, which is used for capturing eye images. We also make our own glasses with the two web-cams settle on it, so as to real-time detect user’s gaze in the scene. Since the glass is very close to user’s eye, we don’t use infrared light sources to help improving our overall accuracy of the system by considering the potential threat to user’s eye under long time infrared illumination. Yet, the performance of our system wins some of infrared illumination works, and we don’t lose much compared with those state-of-art works. In eye blinking detection, we detect one’s voluntary blink get 96.8% overall accuracy on four test subjects, which can be used as a control signal to interact with computers.

並列關鍵字

eye tracking ； Electrooculography ； search coil ； eye loclization ； blink detection ； gaze detection

參考文獻

[1] Andreas Bulling, Daniel Roggen, and Gerhard Troster, “It’s in your eyes: Towards context-awareness and mobile hci using wearable eog goggles,” in Proceedings of the 10th International Conference on Ubiquitous Computing, New York, NY,USA, 2008, UbiComp ’08, pp. 84–93, ACM.

[2] D.A. Robinson, “A method of measuring eye movemnent using a scieral search coil in a magnetic field,” Bio-medical Electronics, IEEE Transactions on, vol. 10,no. 4, pp. 137–145, Oct 1963.

[3] Craig Hennessey, “A single camera eye-gaze tracking system with free head motion,” in In Proceedings of ETRA: Eye Tracking Research and Applications Symposium. 2006, pp. 87–94, ACM Press.

[4] Jixu Chen, Yan Tong, Wayne Gray, and Qiang Ji, “A robust 3d eye gaze tracking system using noise reduction,” in Proceedings of the 2008 Symposium on Eye Tracking Research & Applications, New York, NY, USA, 2008, ETRA ’08,pp. 189–196, ACM.

[5] D.W. Hansen and Qiang Ji, “In the eye of the beholder: A survey of models for eyes and gaze,” Pattern Analysis and Machine Intelligence, IEEE Transactions on, vol. 32, no. 3, pp. 478–500, March 2010.